Tue, Aug 16, 2022: 9:15 AM-9:30 AM
513C
Background/Question/MethodsBiodiversity indicators are essential for providing overviews of biodiversity trends, measuring impact and tracking progress towards conservation targets. The Living Planet Index (LPI) is a global biodiversity indicator calculated as the geometric mean of annual changes in vertebrate species abundance. Within the LPI database, Canadian vertebrates are particularly well-represented as a result of collation efforts undertaken by the Zoological Society of London (developers of the LPI) and WWF-Canada to calculate the Canadian Species Index as well as the national LPI published in the Living Planet Report Canada. Using Canada as a case-study, we investigated whether stratified or random approaches to sampling time-series and species can produce reliable, robust and representative abundance trends for a given country. We aimed to answer the following question. What is the minimum size of a randomly selected sample of species resulting in a trend that is consistent with the overall trend? To address this question, we subsampled the dataset randomly at 10-species intervals with 10,000 iterations per interval creating subsamples with increasing numbers of species. A series of metrics (estimate of the final lambda value, trend trajectory, and field of unexplained variance) were used to compare trends generated from the subsamples with the overall trend.
Results/ConclusionsWe could replicate the direction of the overall trend in 95% of the iterations with a subsample including at least 86% of the total number of species. The median estimated final lambda matched that of the overall trends when at least 75% of the total number of species were sampled. Based on the field of unexplained variance, a similarity to the overall trend (at a 95% confidence threshold) was reached with 77% of the original number of species. Our preliminary results suggest that we can generate a “robust” metric by sampling around 75% of the original number of species in the Canadian LPI dataset. Our next steps will provide an insight on whether a more consistent trend or smaller sample size can be achieved by stratifying the sample by taxonomic group, realm and data-quality (a score based on time-series length and fullness). The interpretation of biodiversity indicators cannot be decoupled from considerations of taxonomic and geographic representation and data quality, all common challenges in monitoring biodiversity. Here we show how those factors interact to influence aggregated biodiversity trends, and hope to help improve representativeness of existing indicators to better assess whether we are on track to meet environmental targets.
Results/ConclusionsWe could replicate the direction of the overall trend in 95% of the iterations with a subsample including at least 86% of the total number of species. The median estimated final lambda matched that of the overall trends when at least 75% of the total number of species were sampled. Based on the field of unexplained variance, a similarity to the overall trend (at a 95% confidence threshold) was reached with 77% of the original number of species. Our preliminary results suggest that we can generate a “robust” metric by sampling around 75% of the original number of species in the Canadian LPI dataset. Our next steps will provide an insight on whether a more consistent trend or smaller sample size can be achieved by stratifying the sample by taxonomic group, realm and data-quality (a score based on time-series length and fullness). The interpretation of biodiversity indicators cannot be decoupled from considerations of taxonomic and geographic representation and data quality, all common challenges in monitoring biodiversity. Here we show how those factors interact to influence aggregated biodiversity trends, and hope to help improve representativeness of existing indicators to better assess whether we are on track to meet environmental targets.